TESI MAGISTRALI

Professoressa De Momi Elena

elena.demomi@polimi.it

Laboratorio: Near lab
Link utilihttp://nearlab.polimi.it/medical/available-master-thesis/                    

The aim of this topic is to improve the manipulability indices of the serial robot during the tele-operated surgery. Based on the achieved teleported MIS demo, the redundancy of the robot arm will be utilized to improve the manipulability of the surgical tip. The experimental evaluation will be conducted on a lab setup environment using KUKA LWR4+ robot and Sigma 7 master device.

Reference:
Jin L, Li S, La H M, et al. Manipulability optimization of redundant manipulators using dynamic neural networks[J]. IEEE Transactions on Industrial Electronics, 2017, 64(6): 4710-4720.

Hang Su (hang.su@polimi.it)

Instead of using fuzzy adaptive control, neural network control will be introduced to work out the uncertain disturbance in the robot workspace using back-stepping design. The aim is to enhance the tool tip accuracy and constrain the RCM constraint error under uncertain physical disturbance in tele-operated surgery. The experimental evaluation will be conducted on a lab setup environment using KUKA LWR4+ robot and Sigma 7 master device.

Reference:
Zhang, Tao, Shuzhi Sam Ge, and Chang Chieh Hang. ?Adaptive neural network control for strict-feedback nonlinear systems using backstepping design.? Automatica 36.12 (2000): 1835-1846.

Instead of using torque sensor, nonlinear observer will be utilized to estimate the extern force for the current Teleported Minimally Invasive surgery. In this way, the controllers do not need torque sensors and can be used for the same tele-operated MIS. The experimental evaluation will be conducted on a lab setup environment shown in the following picture.

Reference:
Sadeghian, H., Villani, L., Keshmiri, M., & Siciliano, B. (2014). Task-space control of robot manipulators with null-space compliance. IEEE Transactions on Robotics, 30(2), 493-506.

Over the last decades, robots applications have started to grow, permeating multiple aspects of everyday life, from industrial manufacturing, to healthcare and automotive. More and more often, humans face the necessity to interact with robotic devices, searching for the best way of transferring human skills to robotic applications.

The proposed thesis fits within the analysis of physical Human Robot Interaction and aims at developing innovative control algorithms to improve cooperative tasks execution performance while interacting with an industrial robotic arm. Users arm dynamic characteristics will be estimated and used to adapt the robotic arm proprieties during the execution of tasks in which the user directly manipulate the robot?s end-effector.

The thesis will provide students with insights in Motor Control theories, real-time estimation of arm kinematics and dynamics, robotic control, EMG signals conditioning and programming. Moreover, machine learning and deep learning techniques could be employed in the development of the novel impedance controller.

Clinical problem: Glioblastoma multiforme is one of the most severe brain tumour with a dramatically poor survival rate of 12-18 months. In the last twenty years, a new surgical technique, convection-enhanced delivery (CED), has shown encouraging results. Its efficacy depends on the ability to predict the drug distribution within the tumour through numerical models. However, most of the constitutive parameters are still unclear.
Objective: This thesis aims to shed light on a fundamental parameter, the hydraulic permeability by means of ex-vivo test on porcine/ovine brain.

Project phases:

Design of a benchmark to test fresh brain tissue (Main target)
Excision of white matter and grey matter samples
Testing and statistical analysis

Brain vasculature visualization is gaining more and more importance in the fields of neuroimaging and neurosurgery. In the specific case of brain vessels segmentation, this task can be highly recommended for the detection of vascular neurological diseases such as ArteroVenous Malformations (AVMs), or for the preoperative planning of neurosurgery treatments such as Stereotactic RadioSurgery (SRS) [3], deep brain stimulation (DBS) implants [4], StereoElectroEncephaloGraphy (SEEG).

At this purpose, we are developing a fully automatic method for the segmentation of arteries and veins obtained through the post-processing of cone-beam CT (CBCT) raw projection data, together with the angiogram obtained from a CBCT digital subtraction angiography (DSA). In this framework, we propose a thesis that aims at investigating new methods for 2D brain vessel segmentation from 2D Digital Subtraction Angiography obtained through CE-CBCT raw data subtraction. This information is crucial for our algorithm and currently it is one of the step that mostly needs to be investigated.

Early-stage diagnosis of laryngeal squamous cell carcinoma (SCC) is of primary importance for lowering patient mortality or after treatment morbidity. Early-stage SCC is associated with the presence of intrapapillary capillary loops and hypertrophic vessels, as well as with the presence of pre-cancerous tissue conditions, such as leukoplakia.

Despite the challenges in diagnosis reported in the clinical literature, few efforts have been invested in computer-assisted diagnosis.

The goal of this thesis is investigating deep-learning strategies to perform fast and accurate SCC diagnosis from endoscopic videos in narrow-band imaging (NBI).

Twin-to-twin transfusion syndrome (TTTS) is a complication of disproportionate blood supply, resulting in high fetus morbidity or mortality. Severe TTTS has a 60?100% mortality rate. Laser therapy is among the main treatments for TTTS. The therapy involves endoscopic surgery using laser to interrupt the vessels that allow exchange of blood between fetuses. However endoscopic field of view is reduced and this poses challenges for surgeons in assessing which vessels have to be interrupted.

On this background, the purpose of this thesis is to investigate new strategies to perform automatic blood-vessel mapping, as to enlarge the field of view and help surgeons in performing laser therapy. The specific goal of this thesis will be investigating stitching strategies for vascular-map composition and deep-learning strategies to automatically highlight vascular structures in the map.

Reference:
Zhang, Tao, Shuzhi Sam Ge, and Chang Chieh Hang. ?Adaptive neural network control for strict-feedback nonlinear systems using backstepping design.? Automatica 36.12 (2000): 1835-1846.

Early-stage diagnosis of laryngeal cancer is of primary importance to reduce patient morbidity. Narrow-band imaging (NBI) endoscopy is commonly used for screening purposes, reducing risks linked to a biopsy but at the cost of some drawbacks, such as large amount of data to review and/or analyze to make the diagnosis. The purpose of this thesis is to investigate new strategies to perform automatic selection of informative endoscopic video frames, as to reduce the amount of data to process and potentially increase diagnosis performance.

Several methods for informative-frame selection have been proposed in the literature, with drawbacks such as heavy parameter tuning and low robustness to inter- and intra-patient variability.

On this background, this thesis will be focused on investigating deep-learning strategies for informative-frame selection, as to provide fast and accurate selection of informative video frames.

On this background, the purpose of this thesis is to investigate new strategies to perform automatic blood-vessel mapping, as to enlarge the field of view and help surgeons in performing laser therapy. The specific goal of this thesis will be investigating stitching strategies for vascular-map composition and deep-learning strategies to automatically highlight vascular structures in the map.

In this way, the controllers do not need torque sensors and can be used for the same tele-operated MIS. The experimental evaluation will be conducted on a lab setup environment shown in the following picture.

Reference:
Sadeghian, H., Villani, L., Keshmiri, M., & Siciliano, B. (2014). Task-space control of robot manipulators with null-space compliance. IEEE Transactions on Robotics, 30(2), 493-506.

Vessel avoidance during nephrectomy is crucial to prevent bleeding and reduce after-treatment mortality or morbidity. In this vision. automatic vessel segmentation can be integrated in robot-assisted surgical procedures as to avoid robot tip to enter (forbidden) vascular regions.

A strong literature already exists on vessel segmentation, however several limitations hamper the translation into the actual clinical practice, such as low robustness to illumination variation and noise, and to variability in patients and imaging instrumentation. Deep learning has the potential to overcome such limitations. Thus, the goal of this thesis is to implement deep-learning based vessel segmentation strategies to perform vessel segmentation in endoscopic videos recorded during nephrectomy.

The proposed thesis fits within the analysis of physical Human Robot Interaction and aims at developing innovative control algorithms to improve cooperative tasks execution performance while interacting with an industrial robotic arm. Users arm dynamic characteristics will be estimated and used to adapt the robotic arm proprieties during the execution of tasks in which the user directly manipulate the robot?s end-effector.

The thesis will provide students with insights in Motor Control theories, real-time estimation of arm kinematics and dynamics, robotic control, EMG signals conditioning and programming. Moreover, machine learning and deep learning techniques could be employed in the development of the novel impedance controller.

Knowledge discovery and data mining is an interdisciplinary area focusing upon methodologies for extracting useful information from data. With advancement of surgeries, there are abundant data generated by intraoperative sensors. The surgery is a process which combines different contextual information e.g. relations between steps, instruments and actions. Inductive logic programming is a data mining technique which could be used to create the knowledge which would be able to help to understand this surgical contextual information. The thesis aims at investigating inductive logic programming methods to understand surgical video, annotated with surgical entities, which could be helpful in context-awareness and decision making.

The current methods of modelling surgical processes are based on top-down modelling approach i.e. modeler?s experience, information in the literature and interviews with the domain experts or any combination of these methods. Top-down approach is highly time consuming, expensive and biased due to human subjectivity. Moreover, surgical processes are dynamic and change constantly, where these process models could not be useful. Data mining techniques could be used to understand the data to discern patterns that would help to develop surgical process model automatically. The proposed thesis will be focused on creating surgical process model automatically from an annotated video dataset of robot-assisted partial nephrectomy, which could be helpful in context-awareness.

At this purpose, we are developing a fully automatic method for the segmentation of arteries and veins obtained through the post-processing of cone-beam CT (CBCT) raw projection data, together with the angiogram obtained from a CBCT digital subtraction angiography (DSA). In this framework, we propose a thesis that aims at investigating new methods for 2D brain vessel segmentation from 2D Digital Subtraction Angiography obtained through CE-CBCT raw data subtraction. This information is crucial for our algorithm and currently it is one of the step that mostly needs to be investigated.

The aim of this topic is, starting from a scaffold of the cerebellar circuit, to implement a cerebellar Spiking Neural Network in the NEST simulator, embedding detailed single neuron dynamics, plasticity mechanisms and geometry-based connectivity; the network will be exploited to simulate sensorimotor paradigms in physiological and pathological conditions.

Project phases:
1. Literature research on the main properties of the cerebellar circuit.

2. Integration of new properties in the NEST-based cerebellar Spiking Neural Network

3.Closed-loop simulations of cerebellum-driven protocols.

Co-Supervisors:
Claudia Casellato (claudia.casellato@polimi.it)
Alberto Antonietti (alberto.antonietti@polimi.it)
Alice Geminiani (alice.geminiani@polimi.it)

Collaborazioni:
Prof. Egidio D?Angelo and Claudia Casellato, University of Pavia
EPFL (Lausanne, Switzerland)

The aim is to develop a classifier of grasping movements based on MYO armband.

Co-Supervisors:
Simona Ferrante (simona.ferrante@polimi.it)
Marta Gandolla (marta.gandolla@polimi.it)

Collaborazioni:
Villa Beretta Rehabilitation Center(http://www.valduce.it/index.php/villa-beretta)

Functional Electrical Stimulation (FES) has been strongly used to improve rehabilitative outcomes after stroke. Hybrid assistive devices, which combines FES with exoskeletons, can maximize the effects of each single technology. Traditional controllers for FES are limited by model complexity, non-linear and time-variant dynamics, and inter and intra-subjects variability.

Project phases:
Starting from a 1-degree of freedom control system based on RF already developed with good results achieved on single target angles:

1. Literature research on control systems for FES
2. Learning of RF control paradigm
3. Design and implementation of a control system able to track a desired angular trajectory and to include the volitional activity of the subject to minimize the FES support
4. Tests on healthy and stroke patients
5. Data analysis of the results

Co-Supervisors:
Emilia Ambrosini (emilia.ambrosini@polimi.it)
Simona Ferrante (simona.ferrante@polimi.it)

Collaborazioni:
Marcello Restelli, Artificial Intelligence and Robotics Laboratory, DEIB, Politecnico di Milano

40% percent of people that suffer a stroke present severe impairments on the upper limb, which affects their quality of life as they are not able to carry out activities of daily living (eating, dressing, cleaning,?). FES has shown several benefits in the neurorehabilitation field with both assistive and therapeutic effects. Surface FES parameters need to be adapted (to each person, to each session,?) and it is critical in the case of the hand due to its neuroanatomical complexity. The aim of the current thesis is the design and development of a closed loop system for adapting stimulation parameters to prono-supination movements of the arm for generating different types of grasps

Project phases:
Starting from available inertial sensor based hand movement sensor system and multi-field FES system for the forearm muscles (extrinsic hand muscles):

1. Literature research on voluntary and FES-assisted hand grasps and closed-loop control techniques
2.Design and execution of data acquisition sessions with healthy subjects
3. Design and development of a closed-loop system for adapting the stimulation parameters to pronosupination movements of the forearm
4. Validation of the approach in lab tests with healthy subjects

Co-Supervisor:
Emilia Ambrosini (emilia.ambrosini@polimi.it)

Collaborazioni:
Thierry Keller, Tecnalia ? San Sebastian (Spain) (thierry.keller@tecnalia.com)

Frailty is a clinically recognizable state of increased vulnerability resulting from aging associated decline in reserve and function. Among other things, it carries an increased risk of falls. Strength and/or resistance exercises as well as training programs based on balance training and/or instability exercises have shown to improve balance and reduce falling risk in elderly and frail populations. Different types of efferent and afferent electrical stimulation protocols have shown to improve spasticity, muscle strength, function and/or balance in elderly and frail populations. The aim of this thesis is the design and integration of a system for its use in an electrical perturbation-based balance training program for frail people.

Project phases:
Starting from available Electrical Stimulation system for the peroneal nerve and the spine extensor/flexor muscles available, pressure platform for base of support measurement and the Equimetrix system for stability index estimation (derived from COM and COP values):

  1. Literature review on balance training programs for improving balance for frail population .
  2. Definition of requirements and specifications of the system.
  3. Design of a communication protocol for integrating the different elements of the system.
  4. Integration of the system.
  5. Validation of the system in lab tests with healthy subjects

Co-Supervisor:
Emilia Ambrosini (emilia.ambrosini@polimi.it)

Collaborazioni:
Thierry Keller, Tecnalia ? San Sebastian (Spain) (thierry.keller@tecnalia.com)

Cycling induced by Functional Electrical Stimulation (FES) has been proposed in the last 10 years to improve lower limb rehabilitation after stroke. Few randomized controlled trials (RCT), usually including a low number of subjects, have been performed. A clear evidence about the effects of FES cycling on motor recovery after stroke is still missing. A RCT (ClinicalTrials.gov #NCT02439515 ) on post-acute stroke patients is currently ongoing at the Institute of Lissone, Salvatore Maugeri IRCCS. The aim of this thesis is to summarize the current evidence about the effects of FES-induced cycling training on motor recovery after stroke, both in terms of a systematic review of the literature and in terms of analysis of the results of the ongoing RCT

Project phases:
Starting from about 60 out of 70 patients that have been already recruited in the ongoing RCT and the corresponding data analysis:
1. Data collection and analysis of the remaining data
2. Statistical analysis of the clinical and instrumental outcome measures and sub-groups analysis
3. Systematic review about the efficacy of FES cycling after stroke
4. Risk of bias assessment and meta-analysis of the included studies
5. Discussion of the results

Co-Supervisors:
Simona Ferrante (simona.ferrante@polimi.it)
Emilia Ambrosini (emilia.ambrosini@polimi.it)

Collaborazioni:
Fondazione Salvatore Maugeri, Istituto di Lissone

Motor system seems to rely on a modular organization (muscle synergies activated in time) to execute different biomechanical tasks. Stroke patients exhibit poor inter-muscular coordination (poor timing and merging of modules that are normally independent in healthy subjects) both during locomotion and cycling, which is correlated to the level of the motor impairment. Different rehabilitative programs might have different effects on motor coordination. This study aims at investigating the effects of a lower limb training on modular muscle coordination during cycling in post-acute stroke patients. Specifically, the effects of a cycling training induced by electrical stimulation will be compared with conventional training.

Project phases:
Starting point: one article has been recently published in Annals of Biomedical Engineering from our group to better understand the neuro-mechanics of recumbent cycling in stroke patients; normative data on healthy older adults are available. Next steps:

1. Literature research on muscle synergies and motor control
2. Data collection on stroke patients
3. Data analysis of EMG and force data pre- and post-intervention
4. Statistical analysis
5. Interpretation and discussions of the results

Co-Supervisor:
Emilia Ambrosini (emilia.ambrosini@polimi.it)

Collaborazioni:
Fondazione Salvatore Maugeri, Istituto di Lissone

Neuroplasticity is an important marker for motor recovery during neurorehabilitation. Transcranial Magnetic Stimulation (TMS) is a non-invasive method to evaluate corticospinal excitability. A rapid method to acquire stimulus-response curves has been recently proposed and its reliability on healthy subjects has been proved. This study aims at investigating the effects of lower limb training on cortical excitability in stroke patients. The final objective is to aid our understanding about how we can augment the effect of motor rehabilitation and identify the optimal treatment plans for its effects to persist and translate to improvements in daily life activities.

Project phases:
Starting from a rapid method to acquire stimulus-response curve already implemented, available normative data on healthy older adults are available and an open source software for manual TMS coil positioning already developed and validated:

1. Literature research on corticospinal excitability in stroke patients
2. Data collection and analysis
3. Statistical analysis
4. Interpretation and discussions of the results

Co-Supervisor:
Emilia Ambrosini (emilia.ambrosini@polimi.it)

Collaborazioni:
Fondazione Salvatore Maugeri, Istituto di Lissone;
Michael Grey,University of East Anglia

Childhood dystonia is a movement disorder characterized by involuntary sustained or intermittent muscle contractions. In case of sensory deficits, children with dystonia may not be aware of their altered patterns of muscle activity and, consequently, they are not able to compensate for unwanted activity. Biofeedback techniques, which provide the subject with augmented task-relevant information, might help improve motor control and accelerate motor learning. A cross-over study has been designed in order to evaluate the effects of a wearable EMG-based biofeedback training in improving motor control in children with dystonia. Both children with primary and secondary dystonia have been recruiting in order to test the hypothesis that the failure of motor learning due to sensory deficits is specific for children with secondary dystonia.

Project phases:
Starting from an EMG-based vibro-tactile Biofeedback device to create awareness of the activity of individual muscles during motor tasks? execution, ethical approval for the cross-over study (first patients have already concluded the protocol):

1. Literature research on biofeedback training in children with primary and secondary dystonia
2. Data collection and analysis
3. Statistical analysis
4. Interpretation and discussions of the results

Co-Supervisor:
Emilia Ambrosini (emilia.ambrosini@polimi.it)

Collaborazioni:
Prof. T. Sanger, SangerLab, University of Southern California, Los Angeles
Dott. G. Zorzi, Istituto Neurologico C. Besta, Milano
Dr. Emilia Biffi, Scientific Institute Eugenio Medea, Bosisio Parini

The improvement of walking abilities is a major goal in the rehabilitation of children affected by neurological impairments. Robot-aided rehabilitation supports traditional methods with some potential advantages including movement repeatability. New platforms integrating treadmills, motion capture systems and virtual reality (VR) offer a more engaging environment. The assessment of the effectiveness of these technologies and the identification of determinants of responsiveness is fundamental to support the choice of the best therapeutic approach

Project phases:
Starting from functional and instrumental data measured before and after robot-aided or VR-aided therapy in children with acquired brain injuries:

1. Literature research on methods to evaluate the determinants of responsiveness to a treatment
2. Development of algorithms to assess gait pattern
3. Assessment of the effectiveness of robot-aided and VR-aided treatment
4. Identification of determinants of responsiveness

Co-Supervisors:
Emilia Ambrosini (emilia.ambrosini@polimi.it)
Emilia Biffi (emilia.biffi@bp.lnf.it)

Collaborazioni:
Dr. Emilia Biffi, Scientific Institute Eugenio Medea, Bosisio Parini

The aim is to identify the longitudinal modifications at cortical level (brain correlates and connectivity), and test/search for predictive markers for the success of FES-based drop foot treatment in post-stroke patients.

Project phases:
1. Brain activity mapping ? i.e. description of the longitudinal modification, if any, in brain activation related to the selected motor task (i.e., ankle dorsiflexion)
2. Connectivity mapping, supervised approach ? i.e. description of the connectivity map and relative longitudinal modification with Dynamic Causal Modeling approach (DCM)
3. Connectivity mapping, non-supervised approach ? i.e. description of the connectivity map and relative longitudinal modification with graph theory approach

Co-Suoervisor:
Marta Gandolla (marta.gandolla@polimi.it)

Collaborazioni:
Dr Nick Ward (UCL Institute of Neurology)

Professor Mainardi Luca

luca.mainardi@polimi.it

Laboratorio: SPINLabS

The research aims at developing methods and algorithms for the
analysis of cardiovascular signals recorded through
non-intrusive devices (smart-watch, wristband devices, etc?)

Application fields:
– Arrhythmia monitoring and classification
– Risk stratification
– Fibrillation (AFib) detection and characterization

The research aims at developing technologies and method for
the measurement of physiological parameters (PPG, HRV and
respiration) using contactless, video-based recordings

Application fields:
– Risk stratification in general population
– Stress quantification
– Neuromarketing

The aim is to simulate the PPG signal during different rhythms (AF, sinus rhythm and other arrhythmias) and to develop new methods to detect the inter-beat intervals. To test the developed method on real PPG data of patients in the
various rhythms.

Data: PPG signals to be simulated (using the model by Solos?enko et al.) and 5-minute real PPG recordings of patients in various rhythms.

Collaborazioni: Lund University; Kaunas University of technology

The aim is to characterize the AV node properties of patients with paroxysmal,
persistent and permanent AF

Data: 5-minute ECG recordings of more than 2000 patients with paroxysmal, persistent and permanent AF part of the Swiss-AF project

Collaborazioni: Swissaf(Swiss Atrial Fibrillation Cohort); Cardiocentroticino

The aim is to characterize the P wave morphology in patients with Brugada syndrome using multi-lead ECG

Data: 10-minute body surface potential mapping ECG recordings of 90 patients with Brugada syndrome, atrial fibrillation and normal sinus rhythm

collaborazione: Universitat Politecnica de Valencia

The aim is to extract Radiomics features from HN lesions and to derive early predictors of the overall survival.

Data: Hundreds of lesions of HN tumors and lymphonode collected during the H2020 Project BD2DEcide

Collaborazioni: BD2DECIDE

The aim is to develop machine learning methods to combine Radiomics and genomic data

Data: Head an Neck lesions concerning tumors and lymphnodes collected since 2004.

Collaborazione: Fondazione IRCCS Istituto Nazionale dei Tumori

The aim is to develop a framework able to highlights these differences on DTI derived maps and their relation with the RT plan on pediatric patients.

Data: 20 pediatric patients undergoing 2 DTI acquisitions (before and after the RT)

Collaborazioni: Fondazione IRCCS Istituto Nazionale dei Tumori; AIRC (Associazione Italiana per la ricerca sul cancro)

The aim is to analyze ventricular Intramyocardial septal Potential using ablation and mapping catheter with a 3D Mapping system

Data: Hundreds of potential coming from catheter in Ventricular ablation procedure

Collaborazioni: Ospedale San Raffaele; Biosense Webster

Collaborazioni:Centro Cardiologico Monzino; Biosense Webster

to analyze ventricular Intramyocardial septal Potential using ablation and mapping catheter with a 3D Mapping system

Data: Hundreds of potential coming from catheter in Ventricular ablation procedure

Collaborazioni: Ospedale San Raffaele; Biosense Webster

Collaborazioni:Centro Cardiologico Monzino; Biosense Webster

Professor Bassi Andrea

andreabassi@polimi.it

La microscopia in vivo di campioni biologici ? una tecnica fondamentale per lo studio del comportamento delle singole cellule nel formare un organismo. Nel caso di interi campioni tridimensionali, ? necessario riconoscere le cellule nell?intero volume, visualizzarne e analizzarne l’evoluzione temporale. Lo studente\la studentessa svolger? la tesi sperimentale in un laboratorio di biofotonica, dove contribuir? alla messa a punto di un sistema laser e all?acquisizione di immagini 3D di organismi multicellulari. Si occuper? quindi di implementare un codice di cell tracking, da poter applicare ai dati raccolti. Le prime settimane di attivit? saranno dedicate alla formazione su sistemi laser, imaging ottico e utilizzo di software (Labview, Python).

Si tratta di un lavoro sperimentale che consiste nella caratterizzazione meccanica mediante prove di trazione biassiale di membrane costituite da uno strato in microfibra di seta ed uno strato in nanofibra. Il lavoro è in collaborazione con una azienda lombarda

Objective of the Thesis is the design and development of an integrated platform to merge and manage data from outcomes measured with clinial tests (such as laboratory tests, medical imaging, ECG, etc) and data from measurements acquired with commercial wearable sensors (such as wristband sensors used for fitness tracking). The purpose is to design such integrated platform to monitor the health status in non-critical patients with chronic diseases and to provide the patient with a detailed and user-friendly journal of his/her health status and the attending doctor a detailed tracking of the patient health status 24 h a day. For the doctor, the platform will provide information useful to assess the fitness/wellness level of the patient and how the daily habits of the patient and the assigned treatment plan influence each other.

Persona di riferimento: Prof. Gabriella Tognola (gabriella.tognola@polimi.it; tel. 3388; Edificio 21)
Collaborazioni con partner esterni: Medas Solutions (Milano, technical partner)

Background: Meniere disease is a disorder of the inner ear that causes severe and debilitating vertigo, hearing loss, and fullness or congestion in the ear. It affects approximately 12 out of every 1,000 people, from 20 to 60 years.
The problem: Although the cause of Meniere disease is well known (buildup of fluid in the labyrinth), no definite answers are available on the many reasons why only some people develop the disease; the treatment options are many , inlcuding medications, surgery, pressure treatments, but the clinical efficacy of these options is not the same for all patients, as it highly depends on the accuracy with which the etiology of the disease has been determined for each single case.
The proposed solution: Objectives of the Thesis are: 1) To develop an automated procedure for clinical text analytics based on Information Extraction able to identify, extract and organize clinical information from the textual medical records that may explain the etiology of the disease (past and present symptoms, medical evidences, signs, past and present treatments, etc); 2) To develop a Clinical Decision Support system that based on the textual information extracted in (1) and on the quantitive (numerical) data from clinical audiometric tests and medical imaging (CT and MR scans) will support the doctor in the differential diagnosis of the disease; 3) To integrate in the System developed in (2) a treatment module to identify clusters of patients with similar treament outcomes and thus support the doctor in planning the best personalised plan for Meniere disease treatment.

Persona di riferimento: Prof. Gabriella Tognola (gabriella.tognola@polimi.it; tel. 3388; Edificio 21).
Partner/collaborazioni esterne: Policlinico di Milano (clinical partner)

Background: Tinnitus (i.e., ringing in the ears) is a symptom that affects nearly 1 in 5 people and may seriously impact the quality of life of sufferers. A range of options are available to mask or distract from the tinnitus, to manage the psychosocial impacts, reduce stress, or to address common comorbidities such as depression, anxiety and insomnia.
The problem: There is a lack of high-level evidence for the clinical efficacy of these different management options. Also, tinnitus-related variables fluctuate substantially over the day, thus it is important to gather the outcomes of tinnitus clinical management in real life (i.e. not in the laboratory) and in real time (i.e. not a single snap shot).
The proposed solution: Objectives of the Thesis are: (1) to develop an innovative platform for collecting real-life and real-time health data through Ecological Momentary Assessment (EMA), and (2) to analyse those variables using Machine Learning to identify a global measure of the outcomes that include but are not restricted to general health, concomitant medications, hearing status, lifestyle, personality, tinnitus-related beliefs, and treatment outcome expectations.

Persona di riferimento: Prof. Gabriella Tognola (gabriella.tognola@polimi.it; tel. 3388; Edificio 21).
Collaborazioni/partner esterni: Policlinico di Milano (clinical partner); Del Bo Tecnologie per L’Ascolto (Milano, clinical partner); Eriksholm Research Centre (Denmark, technological partner)